I would start replacing the two 470µF caps and the TDA chip.

I had the exact same problem on a SBPro2 card and I managed fix it.
I played a 440Hz signal using MPXPlayer in DOS and used a DVM to probe for AC voltage from the Output all the way back to CT1336A on a live card, but I already knew where to probe.
The Input Cap right before the amplifier was dead (discontinuity) and did not pass the signal, but everything before it was the same for both channels.
This problem is quite common with Creative old ISA SB cards for some strange reason...

On your card the input pins to the amplifier should be marked in red (second from top on both sides), so find the caps that go to these pins and replace the correct one:

Thanks ever so much James! I am impressed!

I changed those two caps.. both small 1uF caps on the input pins of the TEA2025. Things are now much better. One channel is producing the level I was expecting and really sounds fine. The other is definitely working (and independent of the first so it's not crosstalk I am hearing) and it's an improvement but still much quieter than needed. I am only using the OPL3 device and not very bothered about the rest btw.

Richard

Post a photo of your mixer settings.
The L/R imbalance is only from the OPL3 part of the card or are PCM sounds (in games or windows player) also imbalanced?

BTW, are you using the Line-Out or Speaker out?
Also, are all the jumpers near the speaker output in place?

Hi James

thanks for all your help!

mixer settings: since I am using the device in a driverless mode the mixer settings are just "write 0 to register 0"

void resetmixer(struct soundcard *sc) {
uint16_t base = sc->baseaddr;
outb(0,base+4); // register 0
outb(0,base+5); // register 0
}

I am using https://pdos.csail.mit.edu/6.828/2011/reading … oundBlaster.pdf as a reference. I have also followed the instructions there to reset the DSP and that works. (Incidentally there are a few oddities, which I put down to me not fully RTFMing when the FM is not available. I also read a yamaha datasheet which didn't tell me very much. I have searched for docs all over the internet for an "official" way to reset the OPL3 on the SB16 and failed. The other major omission in the docs is some sort of timing diagram i.e. how rapidly one can set different OPL3 registers. With modern fast computers this can be rather important. The test computer is a cyrix pentium at 166MHz by the way. But I don't think this is the problem.)

I just tried loading the official sb driver for this particular board with the correct parameters for it. When playing a wave file afterwards, there is no problem with L/R and both channels are equally loud, so the output amplifier paths are all working. (That answers another of your questions.) However, even after loading this driver with the correct parameters the problem is still there using the OPL3. What's more, with all the official drivers loaded I can run one of the standard "playmidi" programs that play a midi file through the OPL3 device. The result is massively weaker sound on one channel than the other, so the problem doesn't appear to be my software at all.

I do not get these problems with other cards I have available to test (an ISAPnP OPL3-SA3 card that needs a driver before I can access the OPL3 ports, an ES1869 PnP virtual card that is on the motherboard)

I just checked the jumpers near the speaker-out port. They seem to be exactly as in your photo. A connection between pins 2 and 3 ( out of 1,2,3,4,5 counting upwards from the bottom ) in the group of 5 nearest the port and the two pins just next to this are connected by another jumper.

To me, unless I missed something, this is pointing to a HW issue between the OPL and the mixer/amplifier...

Richard

sorry, just to add to my last post because I realise I didn't answer all your questions, I get the problem using both line out and speaker out, and... um... games? what are those? windows? nope. This is a minimal linux install. R

Well then, you may have to use Adlib Tracker 2 to generate a test tone and use an oscilloscope to find what pins on the CT1747 output the OPL3 signal and trace the signal to the faulty coupling cap.

yes..

that sounds possible, but it will have to wait. the case prevents any useful access to the rear of the card. I have a dead 486 computer in a much more useful case, but I will have to resurrect that one first... (dead PSU and who knows what else?)

thanks again for your help. if I get any further I will report back